LaunchDarkly
Read and change LaunchDarkly feature flags, segments and the audit trail behind them.
About
LaunchDarkly ships in the w6w first-party pack. It declares 21 actions, 2 health checks, and the host runs its code in a sandbox that never sees the credential.
- App id
io.w6w.launchdarkly- Version
- 0.1.1
- Author
- w6w
- Licence
- MIT
- Categories
- DevOps & Infrastructure · Developer Tools
Overview
LaunchDarkly runs feature flags and the audience segments that target them, and this app puts both under workflow control: read a flag’s state across every environment, turn it on or off, and read or edit the segments its targeting rules point at, alongside the audit trail of who changed what.
Toggling a flag is the headline capability, and it’s treated with the weight that deserves — a change reaches production within seconds, so every write here is logged loudly and takes an optional comment that lands directly in LaunchDarkly’s own audit log rather than only in workflow output. Flags aren’t deleted casually either: archiving is offered as the reversible cleanup step, keeping a flag’s history intact, while permanent deletion needs its own explicit confirmation. A companion action reports whether a flag is still actually being evaluated anywhere, which is the question worth asking before removing one.
Starting or stopping an experiment, and creating new projects, environments or members, stay out of scope — those are decisions with statistical or account-wide consequences that belong with a person, not an unattended workflow step.
Build with LaunchDarkly
Three routes to the same 21 actions. The Workflow tab is generated from LaunchDarkly's own manifest and carries its real ids, so it is copy-pasteable; the Code and CLI examples are the same call for any action on any app, so every app-specific value in them is a blank you fill in.
List audit log entries
audit-log-list Who changed what, account-wide. The only record of why a flag moved.
Get an environment
environment-get One environment and its settings. The response includes SDK keys, which are secrets.
Archive or restore a flag
flag-archive Archive a flag (reversible) or bring one back. Archived flags stop evaluating.
Delete a flag
flag-delete Permanently delete a flag and its history. Archiving is the reversible option.
Get a flag's evaluation status
flag-status-get Whether a flag is still being evaluated — the question before removing one.
List flag evaluation statuses
flag-status-list Every flag's evaluation status in one environment — the flag-cleanup report.
Turn a flag on or off
flag-toggle Switch a feature flag in one environment. Takes effect in seconds.
Update a flag
flag-update Apply semantic patch instructions to a flag — targeting, variations, tags.
Create a segment
segment-create Create a segment in one environment. It does not exist in the others.
Update a segment
segment-update Add or remove segment members, or apply arbitrary semantic patch instructions.
A workflow step names the app and the action, and the editor fills in the
connection when you pick one. This is the Step shape from the
workflow spec, carrying LaunchDarkly's real ids.
{
"manifestVersion": "2",
"name": "launchdarkly-example",
"steps": [
{
"id": "audit-log-get",
"uses": {
"app": "io.w6w.launchdarkly",
"action": "audit-log-get",
"connection": "conn_YOUR_CONNECTION_ID"
},
"with": {
"entryId": "<entryId>"
}
}
]
}Here are some of the things you can do
-
Get an audit log entry
readaudit-log-get -
List audit log entries
readaudit-log-list -
Get an environment
readenvironment-get -
List environments
readenvironment-list -
Create a flag
performflag-create
+16 more actions available
Every app-specific value here is a blank you have to fill in. An
app action is reached through the connection that authenticates it, so the
address is a connection id, not the app id — and connections belong to your account,
so a public page cannot know yours. Create one for LaunchDarkly, then fill in
the three blanks: conn_YOUR_CONNECTION_ID, the action key, and the
parameters that action declares. The call itself is real — the shape is transcribed
from the studio's own snippet builder, which prints the same kind of blanks — but
nothing in it is specific to LaunchDarkly. The Workflow tab is where this app's
real ids are.
npm install @w6w/sdkyarn add @w6w/sdkpnpm add @w6w/sdkdeno add npm:@w6w/sdkimport { W6wClient, isActionRun } from "@w6w/sdk";
// Reads W6W_BASE_URL and W6W_TOKEN from the environment when omitted.
const client = new W6wClient();
const envelope = await client.run({
urn: "conn_YOUR_CONNECTION_ID",
action: "audit-log-get",
payload: {
entryId: "<value>",
},
});
if (isActionRun(envelope)) console.log(envelope.value);npm install -g @w6w/cliw6w run conn_YOUR_CONNECTION_ID --action audit-log-get --payload '{"entryId":"<value>"}'Give an AI agent LaunchDarkly — without giving it LaunchDarkly's credentials. One MCP endpoint exposes every app, function and workflow the caller is entitled to, as tools it can discover and run. Access is granted per team while we onboard.
One tool call{
"name": "w6w_invoke",
"arguments": {
"ref": "app:io.w6w.launchdarkly#audit-log-get",
"input": {
"entryId": "<entryId>"
}
}
}
Every tool names its target with a single ref. The
app: form above doesn't name a connection at all — the
host resolves which of the caller's LaunchDarkly connections to sign
with, and refuses rather than guesses when the answer is ambiguous.
What the agent gets
Credentials it can't read
The token is attached host-side, at the moment of the call. It is never a tool argument, never in the model's context, and never in a transcript — so a prompt injection has nothing to exfiltrate.
A tool surface scoped to the caller
Tools are derived per end user from what that person has actually connected and is entitled to — not one shared bot identity carrying the union of everyone's access.
A durable workflow in one call
Multi-step work runs on the workflow engine and returns a run handle the agent can poll — retries, branching and state survive the conversation that started them.
Health-aware discovery
LaunchDarkly's declared health checks are on the surface too, so an agent can tell "the vendor is down" from "your credential expired" before it burns a retry on either.
The MCP surface is part of the hosted platform. LaunchDarkly itself is MIT, and the runtime that executes it is source-available (FSL).
Health checks
LaunchDarkly declares its own checks, so its health is a property of the app rather than something the host guesses at.
LaunchDarkly platform status
The management components this app calls — authentication, flag targeting, segments, account and audit. Deliberately not the streaming and polling APIs, which are the SDK delivery network.
API request headroom
The global rate-limit headers, if LaunchDarkly sends them. They are documented in prose but declared on no response in the OpenAPI document, so their absence is reported rather than guessed at.